Semiconductor manufacturers have increasingly turned to high density flash memory arrays in their integrated circuit design schemes. The flash memory array includes columns of active regions that are separated by columns of insulating field oxide regions. The transistors are spaced apart in the active regions and each a row of transistors are bits in a memory word. The transistors are formed with various materials including a type-1 layer of polysilicon, and transistors forming a row in the array are connected by a word-line comprising a type-2 layer of polysilicon.
To achieve a high density integrated circuit, the transistors must be as small as possible. Typically, these high density flash memory integrated circuits utilize NAND-type gates as opposed to NOR-type gates since NAND gates have a considerably higher density than NOR gates. Smaller transistors allow more transistors to be placed on a single substrate, thereby allowing relatively large circuit systems to be incorporated on a single, relatively small die area.
FIG. 1 is a top view of a portion of a NAND flash memory array 10. The flash memory array 10 includes a core area 12 and a periphery area 14. The core area 12 includes an array of memory transistors 16 and two select gate regions that include a row of select transistors connected by a select word-line 28. One select gate region is referred to as a select drain gate region 18 and the other select gate region is referred to as a select source gate region 20. Although not shown, the periphery area 14 contains low-voltage transistors for handling logic and switching circuitry, and high-voltage transistors for handling high-voltages encountered during flash memory programming and erase operations.
FIG. 2 is a flow chart illustrating the conventional process steps required to fabricate a flash memory. Also shown is a series of cross sectional views (FIGS. 2(a-f)) of a substrate showing the resulting structure.
The process begins by depositing a layer of tunnel oxide of varying thickness over a substrate in both the core area and periphery areas, via step 30. Next, a layer of type-1 polysilicon (poly1) is deposited in both the core area and periphery area, via step 32. Next, the periphery area is covered and the poly1 is etched from the core area, via step 34. After the poly1 is etched from the core area, a layer of oxide nitride (ONO) is deposited over both the core area and periphery area, via step 36.
After the layer of ONO is deposited, the core area is covered by photo resist and the ONO and poly1 layers are removed in the periphery area, via step 38. Next, a type-2 layer of polysilicon (poly2) is deposited over both the core area and the periphery area, via step 40. Both the core and periphery areas are then etched, via step 42.
The etching process in step 42 is anisotropic, meaning that it removes material directionally to a predetermined depth. But due to the stack height difference between the core area and the periphery area, as illustrated in FIG. 3, the etching process sometimes fails to remove all of the poly1 around the ONO fence area especially at the core/periphery interface and the ONO fence area in the core area, leaving a residue material which is called a stringer. FIG. 4 shows exploded views of FIGS. 2(f) and 2(g) to illustrate the formation of a stringer at the core/periphery interface after the poly2 etch. The presence of a stringer can provide a contact between the two adjacent transistors and failure to remove this material can lead to unwanted electrical shorting paths between the adjacent transistors.
Utilizing the NOR technology, the stringers are not a problem because steps that are implemented later in the NOR process (i.e. dipping steps, oxidation steps), effectively eliminate the stringers. However, as previously mentioned, the NAND process is utilized for high density flash memory integrated circuits since NAND gates have a considerably higher density than NOR gates. Consequently, the NAND process does not incorporate later steps to effectively eliminate the stringers.
Accordingly, what is needed is a method for reducing the occurrence of stringers in the fabrication of flash memory arrays. The present invention addresses such a need.